نبذة مختصرة : Evolutionary algorithms (EA) simulate evolution to solve problems; with each exchange of information and mutation revealing more about the solution space and improving the best solution. These algorithms contain many parameters that need to be set by the user. Chief among these is how each individual in a population is stored, or its representation within the computer. This dissertation proposes a representation that has a simple structure yet generates complex sequences: the self-driving automaton (SDA). This representation will be tested using two problems. The first, network evolution, which aims to generate contact networks that exhibit specific epidemic behaviour when epidemics are simulated. These include: the duration, the number of infections per time step, and the total number of infections of a simulated epidemic. The performance of SDAs is compared to another representation, the edge-editing representation. SDAs are found to excel at exploration of the solution space while the edge-editing representation excels at exploitation within a localized search space. These strengths are combined resulting in a stacked representation which outperforms both of these representations used in isolation. The second problem, sequence matching and pattern discovery, tackles a known problem in biology: identifying a pattern in Φ-segments, sequences found within a class of proteins known as dehydrins. These proteins help plants survive dehydration stresses. Little is known about the Φ-segments, though there is evidence suggesting a pattern may exist within these sequences that could be used to find other dehydrins. The SDAs are first tasked with matching a provided target sequence; achieving 94.8 – 100% matches by encouraging diversity within the population. Next, the SDAs are used to locate a pattern within sets of Φ-segments, which is successful using two fitness metrics. One based on absolute position of DNA characters and the other allowing for gaps which are common in biological sequence matching. Together these ...
Relation: Ashlock, D., & Dube, M. (2021). A Comparison of Novel Representations for Evolving Epidemic Networks. 2021 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 1–8. https://doi.org/10.1109/CIBCB49929.2021.9562847; Dube, M., & Houghten, S. (2022). Now I Know My Alpha, Beta, Gammas: Variants in an Epidemic Scheme. 2022 IEEE Congress on Evolutionary Computation (CEC), 1–8. https://doi.org/10.1109/CEC55065.2022.9870391; Dube, M., & Houghten, S. (2022). Evaluation of Frameworks for Epidemic Variants and Infectivity using an Evolutionary Algorithm. 2022 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 1–9. https://doi.org/10.1109/CIBCB55180.2022.9863031; Dube, M., Sargant, J., & Houghten, S. (2023). Comparison of Representations to Evolve Weighted Contact Networks with Epidemic Properties. 2023 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 1–8. https://doi.org/10.1109/CIBCB56990.2023.10264878; Dube, M., Sargent, J.,& Houghten, S. (2024). Comparing and Stacking Evolutionary Algorithms to Evolve Weighted Contact Networks with Epidemic Properties. BMC Medicine, Submitted for Publication.; Dube, M., Sargant, J., Houghten, S., & Graether, S. (2023). From Bits to Bases: Evolving a Versatile Construct for Biological Sequence and Network Data. 2023 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 1–9. https://doi.org/10.1109/CIBCB56990.2023.10264916; Dube, M., Houghten, S., & Graether, S. P. (2024). Promoting Diversity in the Evolution of Biological Sequence Data. 2024 IEEE Congress on Evolutionary Computation (CEC), 1–7. https://doi.org/10.1109/CEC60901.2024.10611805; https://hdl.handle.net/10214/29059
Processing Request
No Comments.